The legal status of AI-generated images — who owns them, whether they can be copyrighted, and whether training AI systems on copyrighted human artwork constitutes infringement — has become one of the most consequential intellectual property disputes of the current era. In 2025 and 2026, courts in the United States, United Kingdom, and European Union have begun issuing decisions that, collectively, are establishing a legal framework for AI-generated content that will shape the industry for years.
The US Copyright Office has issued the clearest regulatory guidance on AI-generated works. Its February 2023 statement and subsequent clarifications established that copyright protection requires human authorship — a work generated entirely by AI without meaningful human creative control is not eligible for copyright registration. This position was upheld by the US District Court for the District of Columbia in Thaler v. Perlmutter, in which the court ruled that an image generated entirely by an AI system called DABUS could not receive copyright protection because it lacked human authorship.
The more commercially consequential legal question — whether AI systems trained on copyrighted human artwork infringe the copyright of the training data — remains unresolved in US courts. Class action lawsuits filed by artists against Stability AI, Midjourney, and other image generation companies allege that using billions of copyrighted images to train these systems constitutes copyright infringement at scale. The legal theory is that training a model on a copyrighted image is effectively making a copy of it — an act that requires either the copyright holder’s permission or a finding of fair use.
In the UK, the High Court issued a significant 2025 ruling in Getty Images v. Stability AI, finding sufficient grounds for the case to proceed to trial on the question of training data copyright. The EU AI Act, fully enforceable by August 2026, requires AI providers to maintain transparency about training data and comply with copyright law, including honoring opt-out requests from rights holders — a provision that is already reshaping how European AI companies document their training pipelines.
India’s legal framework for AI-generated art is less developed. The Indian Copyright Act does not explicitly address AI-generated works, and no definitive court ruling has established whether such works can be protected or whether training on copyrighted Indian artwork requires licensing. The DPDP Act addresses personal data but does not cover creative works. Legislative clarity on AI and copyright in India is widely anticipated but had not been enacted as of early 2026.
For individual artists, the most practically significant development has been the growth of opt-out registries — databases maintained by AI companies or independent organizations that allow artists to register their work as unavailable for AI training. Adobe Firefly and Getty Generative AI have built models trained exclusively on licensed or public domain content, offering commercial users copyright indemnification as a direct competitive advantage over systems trained on scraped internet data. The market is beginning to segment between provenance-documented, licensed AI art tools and those whose training data sourcing remains opaque.